The observability router calculates estimated costs (using hardcoded $3/1M input + $15/1M output token rates), and the DB schema tracks tokenUsage per agent job. But there is no cost dashboard visible to users, no budget alerts, no per-project spend limits, and no forecast of monthly credit burn. The pricing page mentions credits prominently ('50 AI agent credits / month' on Free, '2,000 on Pro') but the credit system (users.credits field) is separate from actual token usage tracking. Users have no visibility into which agents are most expensive, which products are burning the most credits, or when they're approaching their limit. Competitors like Devin charge per session and show clear cost breakdowns. A dedicated cost intelligence page showing: burn rate by agent type, by product, projected monthly spend, budget threshold alerts, and recommendations for disabling expensive low-value agents would be a direct competitive differentiator in the enterprise segment where finance teams demand cost visibility.